lang-esu

Working Err/Orth tagging of permissive forms

fst: mirror orig analyzer build without Err/Orth tags Build the raw descriptive generator through the permissive phonology path with stress, matching the orig analyzer recipe. Remove local Err/Orth tagging, permissive-only subtraction and disjunction. Apply the ESU remove-boundaries transducer to final analyser/generator products, matching the boundary-removal step in orig.

Add symbolic link pointing to srcV2/ so that the Makefile can do its job

fst: strip ESU enclitic boundary marker from text analysis/generation surface forms The ESU-only "=" morpheme boundary marker was leaking into the surface forms of the descriptive/normative analyser and generator products (analyser-gt-desc/gt-norm, generator-gt-desc/gt-norm), since none of the standard GiellaLT filters know about it. Add filters/remove-enclitic-boundary.regex (a simple "%= -> 0" filter) and compose it on the surface side (to the right of "$<") for exactly those text analysis/generation targets, using a static pattern rule with an explicit target list. This deliberately leaves the raw transducers (analyser-raw-gt-desc, generator-raw-gt-desc) untouched, since they are meant to retain every surface-side boundary marker for use by other applications.

Tag permissive Yup'ik analyses as Err/Orth Build the regular and permissive phonological relations from the shared raw generator, subtract the regular relation from the permissive one, and tag only the remaining paths with +Err/Orth. Mirror ErrOrth flag diacritics on both sides so the tagged distinction survives generator and analyzer processing. Register the tag from the completed raw generator sigma, allowing Giella-core's existing error-tag filters to produce descriptive and normative analyzers and generators without changing srcV2/esu.lexc. Compose lexical cleanup into the lexicon build, include the permissive phonology script, and omit stress marking from the standard morphology pipeline.

Declare %- as part of the alphabet, pr warning from hfst-twolc

Declare missing multi-code-point graphemes

Rebuild with the lang-deps step (divvun-actions#25)

Trigger build: gzipped speller accuracy, optimized accuracy viewer

Rebuild to publish pkg-version.json

Apply changes 343

Trigger build: fst-version.json -> version.json

Trigger build to generate new gramcheck docs data and fix private repo badges

Trigger build: fix lemma test calculation, increase timeout, improve docs testlog reporting

Trigger build: use new rust-wasm typosreport, generate speller accuracy variants

remove missing lexc files

Remove unused typosereport files that now live in jekyll-theme-giellalt

docs: remove committed badge/test data (generated per build now) docs/badgedata/, docs/testlogs/*.md, docs/typosreport/report*.json and the legacy docs/report.json are regenerated on every main build and force-pushed to the generated/docs-data branch, which the README badges and the /testlogs/ and /typosreport/ pages now read live over raw.githubusercontent.com. The tracked copies go stale on every commit and add nothing; .gitignore (rev 335) already keeps them out. Kept: docs/testlogs/index.html and docs/typosreport/{bundle,global,index}.* (page and viewer shells, not data).

Trigger build: fix gtlextools git authentication

Apply changes 334